Seminar: Autonomous structural layout and topology optimization

Igor Melnichenko
Master’s Student
Department of Mechanical Engineering
University of Hawaiʻi at Mānoa

Abstract
This research investigates possibilities of autonomous structural layout and topology optimization method based on cellular division of plants. A genetic algorithm is utilized to generate topology layouts and then FEM software performs sizing optimization for minimizing mass with a stress constraint. This requires a lot of computational power, and therefore supercomputing is utilized. A software package HEDS was worked on during this research. It can generate input files for FEM software ASTROS from a “genome” that is generated by a genetic algorithm of Dakota software. Initial tests were done on a single design generated by HEDS with handmade input files. This design was a simple box fixed at one end with loads applied. Several issues, including singularities, were encountered in results from ASTROS software and were explored by comparing ASTROS results to results from FEniCS Project software. Several modeling solutions were attempted with one being implemented. Tests of the fully implemented method for a box were ran on a personal computer first and then at an AFRL HPC facility. Lighter structures than the baseline were produced, however the layouts are very saturated with certain features. Optimization of an aircraft wing will be performed next. Further improvements of the method can lead to autonomous generation of structural layouts that are lighter than the conventional designs.

About the speaker
Igor Melnichenko received his BS degree in Mechanical Engineering from the University of Hawaiʻi at Mānoa in May 2018. He then worked at Hawaii Evolutionary Development until July 2020. He is currently a MS student and Research Assistant under Dr. Marcelo H. Kobayashi in the Department of Mechanical Engineering.